[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [bmwg] [Sip] Comment on Forking option measurement in draft-poretsky-sip-bench-term-04
Hello Ranjit:
I am redirecting this to sipping since this work is a liaison
between sipping and bmwg WGs.
Avasarala Ranjit-A20990 wrote:
Hi
In the draft: "draft-poretsky-sip-bench-term-04" section 3.3.15 says
that the Forking is an optional process and is not standardized.
Oops, sorry. The "process is not standardized" bit is a cut-and-
paste error from the previous section. It will be taken out
in the next draft revision (which is being worked upon.)
So why is it that the performance of the proxy is impacted by doing
this,
Clearly forking takes up resources in a proxy that would otherwise
not be spent. For instance, in simultaneous forking, the proxy
will have to (if TCP was being used) spend a socket per down-
stream destination, using up kernel buffers for the TCP blocks.
Then, for parallel forking, it has to keep each of the N
transactions in memory, match the incoming responses, figure out
the best response. If it sees a 2xx or 6xx, it will have to
generate CANCELs on N-1 branches, if not it will have to
aggregate Authenticate and WWW-Authenticate headers, etc. All
of this takes up CPU cycles and host resources that would
otherwise be used in servicing other incoming requests.
rather this could be a separate parameter for measuring proxy
performance.
It *is* a separate parameter for measuring proxy performance (if
this is not clear from reading the document, please do let us
know.) The intent is that the bmwg document produces a metric for
performance without forking and with forking, all other variables
being constant between the two runs.
Thanks,
- vijay
--
Vijay K. Gurbani, Bell Laboratories, Alcatel-Lucent
2701 Lucent Lane, Rm. 9F-546, Lisle, Illinois 60532 (USA)
Email: vkg at {alcatel-lucent.com,bell-labs.com,acm.org}
WWW: http://www.alcatel-lucent.com/bell-labs
_______________________________________________
bmwg mailing list
bmwg at ietf.org
https://www.ietf.org/mailman/listinfo/bmwg